Petrologic Investigation of the Lava Creek Tuff, Yellowstone Caldera, WY

Contributions to Mineralogy and Petrology (2019) 174:32 https://doi.org/10.1007/s00410-019-1570-x

ORIGINAL PAPER

Decadal transition from quiescence to supereruption: petrologic investigation of the Lava Creek Tuff, Yellowstone Caldera, WY

Hannah I. Shamloo1 · Christy B. Till1

Received: 19 October 2018 / Accepted: 5 April 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract The magmatic processes responsible for triggering nature’s most destructive eruptions and their associated timescales remain poorly understood. Yellowstone Caldera is a large silicic volcanic system that has had three supereruptions in its 2.1-Ma history, the most recent of which produced the Lava Creek Tuff (LCT) ca. 631 ka. Here we present a petrologic study of the phenocrysts, specifically feldspar and quartz, in LCT ash in order to investigate the timing and potential trigger leading to the LCT eruption. The LCT phenocrysts have resorbed cores, with crystal rims that record slightly elevated temperatures and enrichments in magmaphile elements, such as Ba and Sr in sanidine and Ti in quartz, compared to their crystal cores. Chemical data in conjunction with mineral thermometry, geobarometry, and rhyolite-MELTS modeling suggest the chemical signatures observed in crystal rims were most likely created by the injection of more juvenile silicic magma into the LCT sub-volcanic reservoir, followed by decompression-driven crystal growth. Geothermometry and barometry suggest post- rejuvenation, pre-eruptive temperatures and pressures of 790–815 °C and 80–150 MPa for the LCT magma source. Diffusion modeling utilizing Ba and Sr in sanidine and Ti in quartz in conjunction with crystal growth rates yield conservative estimates of decades to years between rejuvenation and eruption. Thus, we propose rejuvenation as the most likely mechanism to produce the overpressure required to trigger the LCT supereruption in less than a decade.

Keywords Yellowstone · Supereruption · Feldspar-liquid thermometry · TitaniQ · Rhyolite-MELTS · Diffusion chronometry · Sanidine · Quartz

Introduction mechanisms responsible for triggering past eruptions, as well as the timescales associated with specific triggers. Supereruptions, eruptions that produce volcanic products of In recent decades, two end-member models have emerged > 1000 km3 (Sparks 2005), represent an extreme end-mem- to explain the mechanism by which a magma achieves the ber in volcanic eruption style and have the potential to dev- overpressure necessary to drive an eruption: rejuvena- astate human populations globally. Despite the threat associ- tion and volatile exsolution. In the first model hot, lower- ated with these high-risk events, the processes responsible viscosity, crystal-poor melt batches interact with a cooler, for mobilizing large amounts of silicic magma are poorly more evolved, and more crystalline “mush” with varying understood. Improving our knowledge of supereruption pre- degrees of melt, a process commonly referred to as rejuve- eruptive processes requires an understanding of the physical nation (Cashman and Sparks 2013; Bachmann and Huber 2016). For smaller scale eruptions, diffusion chronometry timescales associated with these rejuvenation events are on Communicated by Timothy L. Grove. the order of thousands of years but can be as short as dec- Electronic supplementary material The online version of this ades (e.g., Hawkesworth et al. 2004; Turner and Costa 2007 article (https://doi.org/10.1007/s00410-019-1570-x) contains and citations within; Costa et al. 2011; Till et al. 2015). supplementary material, which is available to authorized users. The second end-member model for achieving overpressure * Hannah I. Shamloo involves volatile exsolution prior to or during magma ascent [email protected] driven by decompression (i.e., “first boiling” Sisson and Bacon 1999) or driven by the crystallization of anhydrous 1 School of Earth and Space Exploration, Arizona State phases (i.e., “second boiling” Sisson and Bacon 1999). For University, 781 Terrace Mall, Tempe, Arizona 85287, USA

Vol.:(0123456789)1 3 32 Page 2 of 18 Contributions to Mineralogy and Petrology (2019) 174:32 example, second boiling has been invoked for sustaining tuff units that precede both members A and B; however, a Plinian column during the formation of the Bishop Tuff their relative volumes remain unknown (Wilson et al. (Wallace et al. 1995). Although, these two models need not 2018). LCT-A and -B are approximately equal in volume be mutually exclusive and are not the only mechanisms that (~ 500 km3) with the ash from the earlier erupted LCT-A trigger eruptions. However, we consider these end members covering a more limited region relative to LCT-B (Fig. 1a). of internal magma body eruption triggers for the purposes of LCT deposits including ignimbrite and ash are distributed this study, as other internal triggers are derivatives of these around Yellowstone caldera and their isopach distributions two models. suggest each member (A and B) erupted from separate cal- In addition to internal eruption triggers, there are exter- dera ring fracture vents (Christiansen 2001). Stratigraphi- nal triggering mechanisms such as magma buoyancy. While cally, members A and B are separated by a layer of fallout thermomechanical modeling suggests magma buoyancy ash from the beginning of the eruption of LCT-B (Fig. 1c), alone cannot provide sufficient overpressure to trigger an the affinity of which is known by the absence of amphi- eruption (< 0.1 MPa; Gregg et al. 2015), it can play an bole found exclusively in LCT-A. Phenocrysts of sanidine, important role in the lead up to an eruption during the rapid quartz, and minor amounts of magnetite, ilmenite, fayalite, assembly of multiple melt pockets, as has been suggested for and zircon characterize the minerology of LCT-A and -B, the Bishop Tuff (e.g., Gualda and Sutton 2016). Other exter- with the addition of amphibole in LCT-A (Hildreth 1981, nal triggers have been investigated including tectonic stress Christiansen 2001). LCT-B fallout ash is the focus of this and magma roof stability (Allan et al. 2012; Cabaniss et al. study, as fallout ash quenches quickly serving as a better 2018; Gregg et al. 2015), although, thus far these processes window into the magmatic processes prior to the eruption. remain difficult to trace in the rock record due to the lack Recent 40Ar/39Ar dating of individual LCT-B sanidine of simultaneous geophysical and geodetic monitoring data. yield LCT eruption ages of 631 ± 4 ka (Matthews et al. Supereruptions have paved the path of calderas known 2015), 630 ± 3 ka (Jicha et al. 2016), and 627 ± 3 ka (Mark as the Yellowstone-Snake River Plain of the western United et al. 2017). U–Pb dating of individual LCT-B zircon faces States. The present manifestation of this continental hotspot, yield crystallization ages of 626 ± 6 ka (Matthews et al. the Yellowstone caldera, has produced three supereruptions 2015), and 629 ± 4 ka from whole-grain isotope-dilution over the last ~ 2.1 m.y., including the Huckleberry Ridge thermal ionization mass spectrometry (Wotzlaw et al. 2015). Tuff, Mesa Falls Tuff, and most recently, Lava Creek Tuff LCT-B zircon are mostly reversely zoned, with zircon faces (LCT). The LCT eruption dispersed ash over the greater and some interior zones having brighter cathodolumines- United States (Fig. 1a, inset), representing a natural labora- cence than their cores with lower U, Th, REE and Hf con- tory for understanding the processes and timescales leading centrations, and smaller europium anomalies, which are to supereruptions. Understanding the lead up to the LCT interpreted as evidence of episodic heating and renewed eruption is also of great interest to both scientists and the crystallization in the LCT source reservoir (Matthews et al. general public due to Yellowstone’s location in the central 2015). Fe–Ti oxide thermometry suggests apparent eruption United States, its on-going hydrothermal and earthquake temperatures from 820 to 900 °C for LCT-B (Hildreth 1981; activity, and its destructive past. This study combines in situ Hildreth et al. 1984) and 819 ± 12 °C from Ti-in-zircon ther- geochemistry via Electron Probe Micro-Analyzer and high- mometry of LCT-B crystal faces (Matthews et al. 2015). spatial resolution Nano-scale Secondary Ion Mass Spec- Wotzlaw et al. (2015) concluded LCT-B zircons record the trometry, diffusion chronometry, feldspar-liquid thermom- rapid assembly of multiple magma reservoirs by repeated etry, TitaniQ thermometry, geobarometry, and petrologic injections of isotopically heterogeneous batches of magma phase equilibria modeling to conduct a holistic, multi-phase with short pre-eruption storage times of 103–104 years. investigation of the events and possible eruption trigger of Lead isotopes measured in individual LCT sanidine the LCT. give 207Pb/206Pb values of 0.88–0.90 and 208Pb/206Pb values of 0.21–0.23, defining a distinct chemical array from Previous studies of the LCT post-caldera rhyolites (Watts et al. 2012). Neodymium and strontium isotopes determined in one study for LCT-B bulk Yellowstone’s most recent caldera forming eruption is tra- are εNd = − 7.8 and 87Sr/86Sr = 0.71012 ± 3 (2σ), which was ditionally thought to have produced two members, LCT-A interpreted as evidence of mafic input into the LCT-B bulk and LCT-B, which represent two eruptive phases indistin- system (Hildreth et al. 1991). Oxygen isotopes measured in guishable in age but separated by a cooling break (Gansecki LCT-B whole rock along with individual zircon and quartz 1998; Gansecki et al. 1998; Christiansen 2001; Matthews grains (δ18O = + 5.3 to + 6.8 ‰) suggest a relatively homog- et al. 2015; Wotzlaw et al. 2015; Jicha et al. 2016; Mark et al. enized and equilibrated magma body prior to the caldera- 2017). A recent study, combining field and geochronology forming eruption with respect to oxygen isotope ratios observations, suggests that LCT may have two additional (Bindeman and Valley, 2001). Glass inclusions in LCT-B

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 3 of 18 32

Fig. 1 a Geologic map of Yel- W111° 00’ W109° 45’ lowstone Caldera (adapted from Matthews et al. 2015). Black A solid lines define the caldera Yellowstone National Park Boundary outline and ignimbrite sheets MONTANA N45° 00’ of both LCT members A and WYOMING B are shown in dark and light gray. Red dot denotes sample location at Flagg Ranch, just outside of Yellowstone National Park. Inset of the United States LCT-B shows ash distribution of the LCT (USGS.gov). b Photograph LCT-A of an outcrop at sample location where the boundary between LCT ash fall and LCT-B ignimbrite are well exposed. c Cartoon showing stratigraphic YELLOWSTONE relationship between ignimbrite CALDERA sheets and fallout ash (adapted Yellowstone from Christiansen 2001; Mat- Lake thews et al. 2015)

Yellowstone

LCT-A N44° 06’ Flagg Ranch Sample location

KILOMETERS 0 20 LCT-B

B C

welded ignimbrite

LCT-B ignimbrite upper LCT-B non- welded ignimbrite ash LCT-B

LCT-B

basal ash upper LCT-A lower LCT-A

1 3 32 Page 4 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

quartz crystals contain H2O concentrations of 3.2–3.8 wt% Sanidine and quartz from the fallout ash were sepa- (av. 3.4 wt%) and CO 2 concentrations of 200–800 ppm (av. rated for within-grain analyses using wet sieving, magnetic 500 ppm) corresponding to pressures of 100–200 MPa in separation, and LST heavy liquids. Euhedral to subhedral the presence of volatile-saturated melt (Gansecki 1998). sanidine and quartz with few to no inclusions were picked These pressures agree with a variety of geophysical depth from the resulting mineral separates for further study. estimates for the uppermost portion of the present-day reser- Phenocryst sizes vary between 500 and 800 µm in width voir beneath Yellowstone of ~ 5 km (~ > 100 MPa pressure) and 1000–1300 µm in height for sanidine, and ~ 1000 by (Eaton et al. 1975; Smith and Braile 1994; Smith et al. 2009; ~ 1000 µm for quartz. Sanidine (135 grains) and quartz (22 Farrell et al. 2014). grains) were mounted in epoxy and imaged by cathodoluminescence (CL) on a scanning electron microscope at Stanford University with a 10 kV accelerating voltage and 20 nA current. CL zoning patterns were used to assess which Methods and samples sanidine and quartz to target for future EPMA analysis (e.g., Fig. 2). For phenocrysts with chemical zoning, we define a Lava Creek Tuff samples “rim” as a bright CL zone with distinct crystal habit that spans from the outermost edge of the crystal to a sharp tran- Bulk ignimbrite and basal fallout ash were collected from sition to a darker CL interior. just outside Yellowstone National Park at Flagg Ranch, WY, where the contact between LCT-A and LCT-B is well Ion‑microprobe major and minor analysis exposed (Fig. 1b). The LCT-B fallout ash deposit (sample of sanidine and quartz 15CTYC-14) is well sorted, containing glassy ash-sized fragments and 17 vol% phenocrysts including sanidine, Compositions of euhedral sanidine and quartz were deter- quartz, and minor amounts of magnetite, ilmenite, fayalite, mined via wave-dispersive spectroscopy using the JXA- and zircon. LCT-B ignimbrite overlies the fallout ash and is 8530F Electron Probe Micro-Analyzer (EPMA) at the non-welded with 88 vol% matrix and 12 vol% phenocrysts Eyring Materials Center at Arizona State University. The including feldspar (8%), dominantly sanidine, quartz (3%), EPMA was operated at 15 kV and 15 nA using a 5 to 10 µm and minor (< 1%) magnetite, ilmenite, ferroaugite, fayalite, beam diameter to measure major feldspar-forming elements zircon, chevkinite, and allanite (sample 15CTYC-17). The plus Ba in 20 sanidine grains (Online Resource 2). Analyses bulk major and trace element compositions of the ignim- of Ti in 4 quartz grains (Online Resource 3) were performed brite and fallout ash were determined by X-ray Fluorescence at 15 kV and 200 nA using a 5 to 10 µm beam diameter (XRF) analysis and Inductively Coupled Plasma Mass Spec- (operating conditions after Thomas et al. 2010). The repro- trometry (ICPMS) at the Washington State University Peter ducibility of standards was typically better than 1 wt% rela- Hooper GeoAnalytical Lab and are given in Online Resource tive for concentrations > 10 wt%, and 10 wt% relative for 1. concentrations of 0.1–1 wt%; thus error is conservatively

Fig. 2 Cathodoluminescent images of representative LCT-B sanidine from each of the three identified populations

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 5 of 18 32 reported as 10 wt% for Ba concentration, 1 mol% for ortho- et al. 2014). The discrepancy in the estimated pressures clase content, and < 0.1 wt% for Ti concentration. may be rooted in the calibrated range of An-contents of the DERP model (3.5–7.0 mol% An; Wilke et al. 2017), whereas NanoSIMS trace element analysis of sanidine LCT-B feldspar has much lower An-content (1.3–2.7 mol% An). Additionally, due to the nature of the model calibration, Based on CL and EPMA data, five representative LCT-B the DERP geothermometer consistently predicts higher pres- sanidine were chosen for trace element analyses with the sures than rhyolite-MELTS (Ghiorso, personal comm., May CAMECA 50L Nano-Scale Secondary Ion Mass Spectrom- 2018). Therefore, for the purpose of this paper, we favor the eter (NanoSIMS) in the NSF SIMS-NanoSIMS Facility at pressures returned by rhyolite-MELTS and use these values Arizona State University. Analyses were made in multi- for further modeling. collection mode using an O − primary beam with Sanidine compositions and thermometry ~ 90–120 pA intensity generating secondary ions with inten- sities of ~ 100–150 nA. Dwell times were 20 s/pixel. A raster size of 256 × 256 µm was used to measure beam-controlled Sanidine from LCT-B fallout ash were grouped into three line scans of 24Mg, 30Si, 88Sr, and 138Ba varying from 40 to populations based on CL-zoning patterns, orthoclase con- 80 µm in length. The ion beam diameter was ~ 0.3 µm with tent, Ba concentrations and their relative estimated tempera- an average of 0.7 µm spacing between each analysis spot. tures. As shown in Fig. 2, Population 1 grains are euhedral Standardization was performed on NIST glass 610, 612, and to subhedral, blocky to tabular, and reversely zoned with 614. Although only the shape of trace element profiles is bright CL rims (20–300 µm thick) that terminate core-ward required for diffusion modeling, absolute concentrations of at a sharp “step-zoned” contact. Overall, Population 1 has measured trace elements were calculated by normalizing to Ba concentrations ranging from 1000 to 7000 ppm with rela- Si and referenced to working curves from NIST glasses. tively low orthoclase content (Or44–54) (Fig. 3). These sani- 2-sigma error on trace element measurements was calculated dine exhibit variable zoning patterns including single thick as Poisson error using the following equation: (130–300 µm) bright CL rims overgrown on a dark CL core, or relatively thin (30–120 µm) bright CL rims that mantle 1∕ Ba cts × Ba (ppm) Si . Temperature (°C) Results 830 825 820 815 810 805 800 7000

LCT‑B geobarometry rim core 6000 Pressure is an important parameter in all the models used in this study; therefore, we determined pressure using two dif- 5000 ferent geobarometers and compared these to previously pub- 4000 lished estimates. First, we utilized the DEtermining Rhyolite Pop. 1 Pressures (DERP) geobarometer, which calculates pressures from the compositions of near-liquidus rhyolitic glass in 3000 equilibrium with quartz and at least one feldspar, consider- Ba (ppm) ing the effect of anorthite and water content (Wilke et al. 2000 2017). Depending on the water content used (1–4 wt% H2O Pop. 2 based on previous LCT work; Gansecki 1998), the DERP- 1000 Pop. 3 calculated pressures for LCT-B vary between 430–460 MPa for the bulk ash (sample 15CTYC-14) or 230–260 MPa for 0 the glass composition (7YC-325, Gansecki 1998). Second, 42 45 48 50 52 55 58 we utilized the rhyolite-MELTS geobarometer (MELTS_ Excel; Gualda et al. 2012; Gualda and Ghiorso 2014), which Or (mol%) yields a pressure range of 80–150 MPa for 1–4 wt% H2O for the LCT-B bulk composition. Rhyolite–MELTS estimates Fig. 3 EPMA data of LCT-B sanidine populations. Population 1 and are consistent with storage pressures determined from recent 2 arrows depict the change in chemistry across core-to-rim bound- melt inclusion data for LCT-B (up to 150 MPa; Befus et al. ary in zoned sanidine. Population 3 data points (grey boxes) are spot analyses across unzoned sanidine grains. Temperature shown here 2018) and the estimated depth of the present crustal magma was calculated using Putirka (2008) feldspar-liquid thermometer with reservoir as determined by seismic tomography (Farrell a standard error of ± 23 °C

1 3 32 Page 6 of 18 Contributions to Mineralogy and Petrology (2019) 174:32 oscillatory zoned interiors (Fig. 2). Crystal rim-core bound- In conjunction with feldspar-liquid thermometry, we aries reveal resorption and regrowth textures where rims utilize alkali-feldspar composition–temperature relation- sometimes truncate preexisting zones with various degrees ships, predicted by rhyolite-MELTS (version 1.1.0), as a of grain boundary rounding. Population 1 is best suited for secondary thermometer. To this aim, we modeled LCT-B further analysis because of their well-defined euhedral edges bulk composition to determine alkali–feldspar–liquid equi- suggesting they record conditions immediately before erup- libria as function of temperature at the appropriate pressures tion. They also represent > 50% of the total sanidine grains (100–300 MPa) over a range of water contents (1–4 wt% inspected and therefore thought to be volumetrically domi- H2O). We then use a linear fit to obtain an equation that nant in LCT-B and thus were investigated in greatest detail predicts temperature from feldspar orthoclase-albite ratio for the purpose of this study. facilitating the determination of temperatures relevant to our Population 2 represents ~ 30% of LCT-B sanidine and is natural LCT-B sanidine compositions where the coexisting characterized by subhedral, tabular grains with relatively melt composition is unknown, for example, for their core thin (30–60 µm) bright CL rims. Some grains have uniform compositions (Fig. 5). Rhyolite-MELTS (Gualda et al. 2012; CL interiors and others have progressive oscillatory zoned Ghiorso and Gualda 2015) has been shown to effectively interiors (Fig. 2). These sanidine have a narrower range in reproduce the crystallization sequence and compositions of their chemical properties compared to Population 1, includ- felsic phases in rhyolitic systems (e.g., the Bishop Tuff), ing lower Ba concentrations ranging from below the detec- particularly the crystallization relationship of sanidine and tion limit (~ 4 ppm) to 2500 ppm as well as have higher quartz (Gardner et al. 2014). Although Gardner et al. (2014) orthoclase content (Or 53–57) (Fig. 3). Population 2 crystals found the absolute temperatures predicted by rhyolite- also show resorption and regrowth textures described for MELTS overestimates absolute crystallization temperatures Population 1. constrained by experiments by ~ 40 °C, in general, rhyolite- Population 3 sanidines are anhedral, mostly fragmented, MELTS and the Putirka (2008) feldspar–liquid thermom- have an orthoclase content similar to Population 1 (Or47–52), eter predicted temperatures within error of each other (e.g., and have the widest range in Ba concentrations of the Fig. 4). Rhyolite-MELTS modeling predicts a temperature three populations, ranging from below the detection limit increase at the core-rim boundary (associated with changes (~ 4 ppm) to 6500 ppm. These sanidine lack both distinct in feldspar composition) of ≥ 10 °C for both Population 1 CL patterns and well-defined crystal rims (Fig. 2) and thus and 2 assuming conditions of 3 wt% H 2O and 100 MPa. were not investigated further in this study. When the orthoclase-albite–temperature relationships for The Putirka (2008) alkali-feldspar-liquid thermometer lower water contents are used (1–2 wt% H2O), the core-rim (reported standard error of ± 23 °C) was applied to the boundaries yield slightly greater temperature changes, up sanidine compositions collected in this study from LCT-B to 20 °C (Fig. 5). Generally, this temperature range would fallout ash in combination with the glass composition from change if crystallization accompanied decompression, which LCT-B ignimbrite pumice (samples 8YC-411 and 7YC-325 is discussed in later sections. Similar to feldspar–liquid ther- reported by Christiansen 2001). The calculations assume mometry temperature estimates, we propagated the analyti- pressures between 80–150 MPa, which represent the calcu- cal uncertainty on major element XRF analyses of the bulk lated pressures from the rhyolite-MELTS geobarometer dis- composition of LCT-B, which yielded ± 35 °C in the abso- cussed above. On average, core to rim changes in orthoclase lute temperature estimates by rhyolite-MELTS. Similar to content are 2–8 mol% for Population 1, and 2–3 mol% for discussion above on feldspar–liquid thermometry, although Population 2 sanidine. Feldspar-liquid thermometry suggests there may be error in the absolute temperatures, we find the an average rim temperature of 814 ± 23 °C at 100 MPa with magnitude of the core-rim temperature variations predicted a temperature increase at the core-rim boundary of ~ 10 °C by rhyolite-MELTS to be robust. for Population 1 and ~ 5 °C for Population 2. Although these changes in temperature are within error of the thermom- Quartz composition and thermometry eter (SEE ± 23 °C; Fig. 4), we interpret these variations as real changes in temperature as they are correlated to varia- LCT-B quartz from fallout ash are euhedral to subhedral tions in orthoclase content determined with very small error grains with reverse zoning in CL. Some grains have rela- (± 1 mol% orthoclase). Furthermore, when the analytical tively thick (100–200 µm) bright CL rims, well-defined crys- error on major element oxide measurements via EPMA anal- tal edges, and sharp boundaries between rims and dark CL ysis was propagated to determine the range in absolute tem- cores (Fig. 6a) similar to the Population 1 sanidine (Fig. 2), perature values, we find very small error of ± 1 °C. There- while others are characterized by oscillatory zoning with fore, although the reported thermometry results for LCT-B thin (25–90 µm) bright CL rims and diffuse intra-crystalline sanidine zones may vary by ± 23 °C, the relative changes boundaries. Overall, LCT-B quartz exhibit resorption and in temperature from crystal core to rim appear to be robust. regrowth textures with rims truncating interior zones with

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 7 of 18 32

Fig. 4 EPMA transects of two Pop. 1 Pop. 2 representative sanidine from Population 1 and 2. Bright resorption CL rims are associated with increases in Ba concentration, decrease in orthoclase content, INTERIOR RIM and slight increase in tem- RIM CORE ZONES perature in both feldspar-liquid thermometry and rhyolite- MELTS thermometry assuming EPMA transect conditions of 3 wt% H 2O and 100 MPa. Temperature esti- 400 μm mates from feldspar-liquid ther- 57 mometry and rhyolite-MELTS 56 show absolute temperature esti- 55 mates within error [SEE ± 23 °C Pop. 1 (Putirka 2008), ± 35 °C rhyolite- 54 Pop. 2 MELTS] but still reflect real 53 core-rim temperature fluctuations based 52 on orthoclase variation (see text boundary for further discussion). Popula- 51 tion 1 sanidine have greater 50 core-rim variation in these properties Orthoclase (mol%) 49 boundary from core to rim compared to Population 2 grains. The error 48 on Ba is smaller than symbol 47 size in some instances 825 840 core-rim boundary 820 828 (°C) (°C) 815 core-rim 816 fsp-liq boundary

810 804 rhyolite-MELTS

Temp. 805 792 Temp.

800 780 7000 core-rim 6000 boundary 5000

4000 core-rim 3000 boundary Ba (ppm) 2000

1000

0 50 100 150 200 250 300 350 distance from rim (µm) various degrees of rounding, similar to Populations 1 and 2 Titanium-in-quartz thermometry (TitaniQ; Wark and sanidine. EPMA analysis shows an average increase in Ti Watson 2006) for LCT-B quartz rims was conducted over concentration from core to rim of 20 ppm, with average rim a pressure range of 80–150 MPa consistent with the geo- a concentrations of 40–80 ppm Ti. barometry discussed above, at a constant TiO2 of ~ 0.33 as

1 3 32 Page 8 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

840 change from core to rim remains the same regardless of the 1 wt% H2O 2 wt% H O chosen activity and pressure. The average temperature for 830 2 quartz rims varies from 760–850 °C ± 35 °C (av. 815 °C) 820 corresponding to the range of activities and propagated ana- Putirka (2008) a lytical error via EPMA. Although TiO2 is assumed constant 810 for these calculations, which may not be the case in nature, 3 wt% H O 2 a temperature increase from quartz core to rim is still resolv- 800 able. For example, in the hypothetical scenario where a reju- 790 venation event is accompanied by remelting of the resident emperature (°C)

T a a LCT-B reservoir, where TiO2 increases (e.g., TiO2 = 0.33 780 a to TiO2 = 0.6, the total range of Yellowstone lavas reported by Vazquez et al. (2009)), would correspond to an absolute 770 temperature change of only 8 °C in TitaniQ. Thus, even in 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 this scenario, changes in quartz composition correspond to Orthoclase/albite temperature increases > 20 °C. Modeling LCT‑B bulk system to constrain Fig. 5 Rhyolite-MELTS predicted relationships between temperature pre‑eruptive conditions and orthoclase–albite ratio assuming 100 MPa and variable water contents. Each trend is linearly fit and used as a thermometer to predict temperatures for LCT-B sanidine orthoclase–albite ratios. Grey The sequence of crystallization for LCT-B bulk ash compo- thick line represents LCT-B (Population 1 and 2) sanidine core and sition was determined using rhyolite-MELTS (Gualda et al. rim compositions with temperatures predicted by feldspar-liquid ther- 2012; Ghiorso and Gualda 2015). For these compositions, mometry (standard error of ± 23 °C not shown; Putirka 2008) the ferrous–ferric ratios were determined via MagmaSat and isobaric crystallization was modeled between 1250 and suggested by other Yellowstone lavas similar in major ele- 500 °C. Isenthalpic mode (− 50 J) allowed for better deter- ment chemistry as LCT-B (e.g., Solfatara flow, Vazquez et al. mination of the ternary minimum and greater resolution in 2009). Here we use the Huang and Audétat (2012) TitaniQ temperature as this mode allows the temperature to vary at calibration, wherein the activity term is included in the equa- each step along the calculation path (Gualda et al. 2012; tion, as indicated by Hayden and Watson (2007), and returns Ghiorso and Gualda 2015). The models were run with a a relative temperature increase across rim-core boundaries of range of initial water contents (1–4 wt% H2O) and a CO2 30–50 °C. Although TitaniQ requires important assumptions concentration of 500 ppm suggested by melt inclusion data to determine absolute temperature, the relative temperature for LCT-B ash (Gansecki 1998). Modeling at 100 MPa and

A quartz B (860),94 130 EPMA core-rim transect rim boundary (840), 84 120

ImageJ (820), 74 110 , Ti (ppm) transect core (800), 64 100

(780), 54 90 Grey scale

(760), 44 EPMA measurements 80 CL intensity Temperature (°C) (740), 34 70 0 0 5 25 50 75 10 12 150 400 μm distance from rim (µm)

Fig. 6 a Cathodoluminescent images of LCT-B quartz grains with niQ temperatures are also shown (values in parenthesis) with a rela- variable zoning patterns. Location of a longer EPMA transect (red) tive temperature increase towards the outermost rim. Error bars repre- and shorter ImageJ transect (yellow) are included. b Ti concentration sent ± 5 ppm Ti per the reproducibility of standards on EPMA. Error profile analyzed via EPMA (black circles) overlaid by CL intensity from Ti measurements and variation in activity were used to assess profile (grey triangles) extracted from a CL image using ImageJ soft- error on temperature calculations of ± 35 °C (not shown, see text for ware, illustrating the correlation between Ti concentration and CL discussion) intensity similar to what is observed by Leeman et al. (2008). Tita-

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 9 of 18 32

830 39

T sanidine 43 35 0.8

im 47 r

vg. quartz 0.6 800 51 B. 2. solidus A. and qtz a 3. albite C. 1. san. 770 magnetite 0.4 H O+ CO opx 2 2 rhm-oxides 55 F = 0.2 feldspar 740 emperature (°C) T 710 59 Fig. 7 Phase diagram of the bulk LCT-B system predicted by rhyolite-MELTS modeling at 3 wt% H2O under isenthalpic and isobaric Or = 63 conditions of 100 MPa. Sanidine and quartz rim temperatures calcu- 680 lated via thermometry are shown by dashed line 50 100 150 200 250 300 350 400 Pressure (MPa) 3 wt% H2O, as suggested by both geobarometry and melt inclusion estimates discussed above, produces a crystal- Fig. 8 Phase diagram showing the stability of variable orthoclase lization sequence with magnetite as the liquidus-phase at using rhyolite-MELTS predicted pressure-temperature-composi- 920 °C, followed by feldspar at 830 °C, quartz at 820 °C, tion data for the LCT-B bulk composition at 3 wt% H2O. Blue con- and the feldspar solvus at 820 °C, below which sanidine and tour lines represent orthoclase content, and grey dashed lines are melt albite crystallize under fluid-saturated conditions (Fig. 7). fraction contours. Two sets of black arrows are shown as trajectories showing three major possibilities to achieve the orthoclase change Although the crystallization conditions of mafic phases are from core to rim observed in LCT-B sanidine (av. Or52 to Or 49) if less reliable in rhyolite-MELTS (Gardner et al. 2014), the assuming a starting pressure of 100 MPa (A–B) or 300 MPa (1–3), model predicts orthopyroxene and rhombohedral-oxide crys- for example. Note we favor the set of trajectories at 100 MPa, as it tallization at 810 °C and 740 °C, respectively. At 814 °C, corresponds to geobarometry estimates the average temperature of LCT-B sanidine and quartz rims, the modeled bulk system consists of 85% melt, 6% sanidine, the average change from Or 52 cores to Or 49 rims observed 4% quartz, 3% albite, 0.8% magnetite, 0.2% orthopyroxene, in the LCT-B Population 1 sanidine. Two sets of trajecto- and 1% fluid (48 wt% CO2: 52 wt% H 2O) (Figs. 7, 8). This ries are shown assuming starting pressures of 100 MPa and crystal assemblage closely reflects the modal abundance of 300 MPa. Note that we favor 100 MPa per the geobarometry the natural ash sample (sample 15CTYC-17). Over the tem- results explained in previous section. The P–T trajectories perature interval relevant to both LCT-B sanidine and quartz that reproduce the observed change in sanidine composition predicted by thermometry, rhyolite-MELTS supports that from core to rim include: (1) decompression of 50–100 MPa sanidine and quartz are co-saturated, a phenomenon com- during heating, (2) an isobaric heating event or, and (3) an monly observed in the haplogranite ternary system (e.g., increase in pressure of ~ 25–50 MPa during heating. Gualda and Ghiorso 2014; Bachmann and Huber 2016). Therefore, we invoke that LCT-B sanidine and quartz repre- Timescales via diffusion modeling sent two major phases indicating crystallization in a system that was simultaneously reheated. Diffusion chronometry utilizing NanoSIMS Ba and Sr con- In addition to modeling the sequence of crystallization, centration profiles in sanidine was conducted on five rep- a series of successive isobaric crystallization models (runs resentative crystals from Population 1 (Online Resource over a conservative pressure range of 50–400 MPa) were 4). Changes in Ba concentration from core to rim were on compiled to construct a phase diagram (Fig. 8). Feldspar the order of 2000–4000 ppm for Population 1 grains and orthoclase content (solid blue lines) and bulk melt fraction 1000 ppm for those in Population 2 (Fig. 4). Diffusion chro- (dashed gray lines) are used to determine pressure–tempera- nometry was also conducted on LCT-B quartz grains; how- ture (P–T) trajectories that could reproduce the observed ever, analysis of Ti concentration in quartz is challenging major element zoning in LCT-B sanidine. The results from due to its relatively low abundance and limited spatial-res- rhyolite-MELTS and the derivative phase diagram suggest, olution on the EPMA. Like many other studies (e.g., Wark in all cases, an increase of ~ 10 °C is necessary to produce and Watson 2006; Wark et al. 2007; Matthews et al. 2012;

1 3 32 Page 10 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

Leeman et al. 2012; Gualda et al. 2012; Gualda and Sutton average timescales on the order of ~ 0.02 ky (Fig. 9, Online 2016), we observed a correlation between Ti concentration Resource 4). If produced by the same event, the Ba and Sr and CL intensity (Fig. 6b), and thus utilized CL intensity profiles should yield similar timescales despite their differ- as a quantitative proxy for Ti concentration in quartz. CL ent diffusivities. Additionally, because Sr diffuses faster than intensity profiles were extracted using ImageJ (Schneider Ba, the width of the Sr profile should be greater than that of et al. 2012) image processing software on six LCT-B quartz Ba for a given time interval (Table 1). However, in almost all grains (Online Resource 5). cases the Ba and Sr profiles are of equivalent width and in Modeling assumed a step-function initial condition and one case, the Sr profile is shorter than the Ba profile (Online employed an analytical solution to Fick’s Second Law, Resource 4), which is the opposite of what is expected. This involving one-dimensional diffusion in an infinite medium suggests that the profiles did not originate as step-function with an abrupt change in composition when the diffusion concentration profiles subsequently altered by only diffusive distance is small (Costa et al. 2011): relaxation. Till et al. (2015) observed similar behavior in Ba and Sr concentration profiles across sanidine core-rim C1 − C0 x boundaries from a post-caldera rhyolite at Yellowstone and C = C0 + erfc , (1) 2 � 2 Dt � concluded the profile originated during rim growth in an evolving melt composition with little to no subsequent diffu- √ sion. Here we draw the same conclusion, which is also sup- where C is the normalized concentration, C0 and C1 are the initial amounts of the elements on each side of the interface ported by the relative rates of diffusion versus crystal growth −8 −12 −1 at time zero, D is the diffusivity(m 2 s−1), t is the diffusion in sanidine [i.e., sanidine growth rates (10 –10 m s ; time, and x is the midpoint on the concentration profile. Diffusivities were calculated using Eq. (2) using the Arrhe- 1050 nius parameters determined by Cherniak (1996, 2002), and data Cherniak et al. (2007) specified in Table 1. For sanidine, rim best ﬁt modeling assumed a constant temperature of 814 °C repre- 1000 senting the average rim temperature as determined by feldspar–liquid and rhyolite-MELTS thermometry. For quartz, the modeling was conducted at a range of temperatures 950

between 760–850 °C, representing quartz rim temperatures Ba (ppm) determined by TitaniQ specified above. core 900 −E D = D e RT . (2) model time = 1 ky 0 850 We used a least-square minimization procedure to find µ 59 the complementary error function that best fit each observed NanoSIMS transect trace element profile by varying the concentration of each 56 plateau, the center of the diffusion profile, and the diffusion length scale ( Dt). 54 rim CORE When modeled√ at a constant temperature of 814 °C, RIM the Ba in sanidine profiles suggest average timescales on 52 the order of 1 ky, while the Sr in sanidine profiles suggest Sr (ppm ) 400 μm 49 core 46 model time = 0.02 ky Table 1 Arrhenius parameters and calculated diffusivities 44 2 D0 (m /s) E (kJ/mol) Diffusivity at Profile width at 814 °C (m2/s) 814 °C for 5 years 22 23 24 25 26 27 28 29 30 31 32 (μm) distance (µm)

Sanidine a Fig. 9 NanoSIMS trace element profiles across core-rim boundary for Ba 0.29 455 3.9E−23 0.5 b a representative Population 1 sanidine. The best fit model calculated Sr 8.4 450 2.0E−21 3.5 via diffusion chronometry is shown in red at an assumed temperature Quartz of 814 °C with a step-function as an initial condition. NanoSIMS Tic 7.0E−08 273 5.3E−21 6.0 transect (yellow) direction is from rim (left) to core (right) as shown on CL image provided. Error bars are smaller than symbol size in the a Cherniak (2002), bCherniak (1996), cCherniak et al. (2007) case of Ba measurements

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 11 of 18 32

Swanson 1974, 1977) are far faster than Ba and Sr diffusion 180 (10 −22–10−23 m2 s−1; Table 1)]. In this situation, any time- 1600 150 scale calculated via diffusion modeling of Ba or Sr relaxa- 1200 800 120 tion from a step function will represent a significant over- 400 estimation of the time it took for the rim to form. Instead, 0 100 0 0 diffusion chronometry requires the comparison of the Ba 750 770 79 81 830 850 and Sr concentration profiles with that of a very slow dif- 80

fusing element at magmatic conditions, such as CaAl–NaSi Time (yr) 50 exchange in plagioclase that can be used as a proxy for an initial profile shape prior to diffusion. However, at present, 30 we lack such data on similarly slow diffusing elements for sanidine. Because the Ba and Sr concentration profiles are 0 0 identical in shape and width in four of five grains modeled, 750 760 770 780 790 800 810 820 83 840 850 we instead constrain a maximum possible time interval Temperature (°C) between sanidine rim growth and eruption by determining the time required for the faster diffusing Sr to diverge in Fig. 10 shape from the Ba profile using a finite difference approach The range of possible timescales at a given crystallization temperature for LCT-B quartz rims modeled over a range of pres- a similar to Till et al. (2015). This approach yields maximum sures (100–300 MPa) assuming a constant TiO2 of 0.33 (Vazquez timescales of 10–20 years between rim growth and eruption et al., 2009). Symbols represent the rim temperatures determined for for the LCT-B sanidine at 814 °C. Therefore, we infer the one quartz grain. Lines are linear fits for the range of timescales and Population 1 high-Ba and -Sr sanidine rims formed less than temperatures for a single grain to illustrate the relationship between time and temperature-dependent diffusion. Inset shows an outlier 10–20 years prior to the eruption. grain with relatively long timescales not shown in the range of the For quartz diffusion modeling, a step-function initial main plot condition was assumed in the absence of comparison to other trace elements with varying diffusivities. We utilize 790–815 °C for the average quartz rim growth temperature assume a constant growth rate or linear behavior, where tem- during modeling as it corresponds to the preferred pressure perature and pressure are kept constant, and growth is there- range of 80–150 MPa from geobarometry and is within error fore proportional to time (Zhang 2008). In nature, however, of the sanidine rim thermometry. Modeling at average rim this is likely not the case. Therefore, given the uncertainty temperature 815 °C yields timescales of 3–40 years, and in experimentally derived crystal growth rates applied to in one case 300 years (Fig. 10, Online Resource 5). With- natural systems, we favor the timescales predicted by dif- out the concentration profiles for elements with differing, fusion chronometry for sanidine and quartz. Thus, the data well-constrained diffusivities, we take these timescales to suggest that the LCT sanidine and quartz rims grew less than represent maxima. a couple of decades prior to eruption.

Timescales via crystal growth rates Discussion In the absence of a faster diffusing element in sanidine and a second element with different diffusivity in quartz, we LCT‑B magmatic history reconstructed also examined crystal growth rate estimates for feldspar through the phenocryst record and quartz in rhyolitic melts to compare to the maximum timescales determined through diffusion chronometry. Crys- The chemical zoning patterns and diffusion chronometry tal growth rates from both experimental observations and of volcanic phenocrysts are increasingly used as a record natural samples for felsic magmas, under pressure and tem- of the changes in magma’s history preceding eruption, but perature conditions relevant to this study, suggest a range rarely is a distinct triggering mechanism responsible for an between 10 −8 and 10−12 m s−1 for both feldspar (Swanson eruption thoroughly resolved. Before assessing plausible 1974, 1977; Fenn 1977; Zellmer and Clavero 2006; Cal- eruption triggers, however, we must first constrain the pro- zolaio et al. 2010), and quartz (Swanson 1974, 1977; Gualda cesses recorded in the LCT-B phenocrysts. LCT-B sanidine and Sutton 2016). Assuming the slowest growth rate for both and quartz both possess resorption textures at their core- feldspar and quartz (10−12 m s−1) produces conservative esti- rim boundaries and reverse-zoning with bright CL rims with mates of 7 months to 20 years (av. 5 years) for the growth of elevated Ba and Sr concentrations in the case of sanidine, LCT-B feldspar rims, and 7 months to 4 years (av. 2 years) and elevated Ti in the case of quartz. Both phases also record for LCT-B quartz rims. These calculated growth timescales a ubiquitous increase in rim crystallization temperatures

1 3 32 Page 12 of 18 Contributions to Mineralogy and Petrology (2019) 174:32 compared to their interiors. Together these phenocrysts sug- in LCT-B quartz rims and could explain the elevated Ti gest a composite history including the following sequence concentrations in the quartz rims. The other scenarios are of events; (1) the growth of phenocryst cores (which are explored further below. sometimes oscillatory, suggesting complex early histories), (2) subsequent heating and crystal resorption, followed by Decompression (3) growth of sanidine and quartz rims with elevated magmaphile element concentrations (Ba, Sr, Ti) in decades or Decompression is often accompanied by an increase in tem- less prior to eruption, as suggested by diffusion chronometry perature (e.g., Figure 8, scenario A/1) due to the latent heat and growth rates (Figs. 3, 4, 6, 8, 9). In the following sec- of crystallization (e.g., Blundy et al. 2006). For example, tion we interrogate a series of possible magmatic processes for the case of a slowly ascending H2O-saturated rhyolite responsible for the LCT-B phenocryst record, a number of melt, thermodynamic calculations suggest temperature which are motivated by the P–T paths necessary to produce increases on the order of 25 °C for a 200 MPa pressure the LCT-B sanidine shown in Fig. 8. These possibilities drop, or 2.3 °C per 1% of plagioclase crystallization (Couch include (a) fractional crystallization, (b) decompression, (c) et al. 2003; Blundy et al. 2006). The maximum change in the addition of volatiles into the magmatic system, and/or pressure consistent with the observed decrease in orthoclase (d) a replenishment of less-evolved material into the system content in LCT-B sanidine rims (150 → 50 MPa: Fig. 8, sce- (e.g., rejuvenation or underplating event), as well as related nario 1/A), would thus produce an increase in temperature thermal and decompression-induced changes in elemental of ~ 10–20 °C. This pressure drop is also consistent with partitioning. the temperature change suggested by both rhyolite-MELTS and feldspar-liquid thermometry. The same logic follows for Fractional crystallization and trace element partitioning LCT-B quartz. TitaniQ thermometry suggests a temperature increase from crystal core to rim of 30–50 °C (± 35 °C), The bulk partition coefficients for Ba and Sr for LCT-B and like LCT-B sanidine, latent heat of crystallization can whole rock composition are controlled by their relative account for ~ 10–20 °C or potentially more. Given its explo- compatibility in feldspar and quartz, the two most abun- sive nature, it is likely that the ascent rate for the LCT-B ash dant phases (60%, 30%) of the LCT phenocryst assemblage. was much quicker than what was invoked in Couch et al. Thus, progressive fractionation of sanidine from the host (2003) and Blundy et al. (2006), such that it could account magma would decrease Ba and Sr concentrations in the sani- for > 20 °C of heating in some situations. dine from core to rim (the opposite of which is observed). Decompression would also result in an increase in Ti-in- Furthermore, Ba and Sr feldspar–liquid partitioning is not quartz partitioning, which is difficult to distinguish from an temperature- or pressure-dependent. Rather, the partition- increase in magmatic temperatures (Wark and Watson 2006; ing of Ba and Sr in alkali-feldspar has been shown to be Thomas et al. 2010; Huang and Audétat 2012), similar to heavily dependent on major element feldspar chemistry or what was recently invoked to explain high-Ti quartz rims orthoclase content for silicic melt compositions (Long 1978; formed within a year of eruption invoked for the Bishop Smith and Brown 1988; Guo and Green 1989; Blundy and Tuff (Gualda and Sutton 2016). Thus, decompression- Wood 1991; Icenhower and London 1996; Ren 2004). There- driven crystallization (combined with pressure-dependent fore, the decrease in orthoclase content within LCT-B rims changes partitioning for the quartz) explains the chemical (Fig. 3), should be accompanied by a decrease in Ba and Sr and thermal profiles recorded by LCT-B sanidine and quartz, content. Again, this is the opposite of what is observed, sug- although it does not fully explain the resorption textures gesting fractional crystallization and changes in partitioning observed between the phenocryst cores and rims. are not responsible for the formation of the chemical profiles observed in LCT-B sanidine rims. Addition of CO 2 Similarly, LCT-B quartz rims are enriched in Ti. This is a common observation for quartz phenocrysts in silicic mag- Another mechanism possible to facilitate crystal growth mas and Ti-rich rims overgrown on Ti-poor cores have been and the Ti-enrichment in quartz includes the interaction of interpreted to reflect either (1) a temperature increase (e.g., CO2-rich vapors from a deeper and less-evolved body with Müller et al. 2003; Wark et al. 2007), (2) a pressure decrease shallowly stored magma (e.g., Blundy et al. 2010; Oppen- (Wark and Watson 2006; Cherniak et al. 2007; Thomas et al. heimer et al. 2011; Iacovino 2015; Humphreys et al. 2016). 2010; Huang and Audétat 2012; Gualda and Sutton 2016), There is some evidence to suggest CO2 is an important (3) the addition of CO 2 into the magma (Gansecki 1998; player in the LCT source region, where the injection of a Wark et al. 2007; Blundy et al. 2010) and/or (4) the build- less-evolved magma either accompanied or followed iso- up of a boundary layer in the melt followed by rapid growth baric fluid-saturated crystallization (e.g., Fig. 8, scenario (Pamukcu et al. 2016). A temperature increase is observed B/2), a mechanism invoked to explain the variability of

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 13 of 18 32

CO2 (100–760 ppm) compared to H 2O (3.2–3.8 wt%) in change in chemistry recorded by LCT-B phenocryst rims quartz-hosted melt inclusions in LCT-B fallout ash (Gan- is not significant enough to require the addition of a mafic secki 1998). The rhyolite-MELTS modeling in this study magma, rather; it is more likely that a magma body with that best matches the LCT-B phenocryst rim thermometry similar major element chemistry (i.e., rhyolitic) but with also suggests that LCT-B sanidine and quartz rims were higher concentrations of compatible trace elements (rela- crystallizing in the presence of a fluid rich in CO2 (Fig. 7). tive to LCT-B cores), which is the same scenario interpreted While the addition of CO2 could act as the driving force for for post-LCT rhyolite lavas based on their Hf and O isotope crystallization of LCT-B phenocryst rims and supply the compositions in zircon (Stelten et al. 2013, 2015, 2017). temperature increase through the latent heat of crystallization, this mechanism alone cannot explain the trace element Implications of phenocryst populations distribution in sanidine and quartz (e.g., Figs. 4, 6) based on the associated partitioning behavior previously discussed. Recall, there are three populations of LCT-B sanidine, as Rejuvenation and/or underplating well as different zoning patterns in LCT-B quartz (Figs. 2, 6a). We interpret the Population 1 sanidine as recording a single recharge event, while the Population 2 sanidine record Even in the most favorable scenario discussed thus far, multiple small events. The same is implied from the vari- decompression alone fails to explain the resorption textures able zoning patterns recorded in LCT-B quartz. The unzoned observed between the cores and rims in LCT-B sanidine and Population 3 sanidine have orthoclase and Ba contents com- quartz. Underplating of more juvenile material would supply parable to both the core and rim compositions of the Popula- heat and potentially melting to the LCT-B reservoir, such tion 1 sanidine (Fig. 3). One possibility is that the Ba-poor that resorption textures could develop on the phenocryst Population 3 crystals were derived from the magma in the cores, and enrich the melt in magmaphile elements (Ba, Sr, system that did not mix with the recharge magma, while the Ti). However, at the relevant temperatures, Ba and Sr would Ba-rich Population 3 crystals were derived from the recharge diffuse 200 and 900 µm in 20 years, exceeding the aver- magma itself, as dynamic simulations of mush systems sug- age Population 1 rim thickness (160 µm), precluding the gest individual crystals see dramatically different P–T–t likely buildup of an enriched melt-layer around feldspar as histories during a rejuvenation event depending on their Pamukcu et al. (2016) suggested. Melting crystals via under- location (Wallace and Bergantz 2002; Bergantz et al. 2017; plating could still enrich the ambient melt in Ba, Sr, and Schleicher and Bergantz 2017). Alternatively, the high-Ba Ti; however, it is difficult to know by how much. Thus, we Population 3 sanidine may have crystallized in a homog- cannot rule out underplating, although it is not our preferred enized melt composition after mixing and shortly before or mechanism for providing the trace element enrichment in concurrent with the explosive eruption of the LCT-B (e.g., the phenocryst rims. Streck 2008 and references within). Overall, we interpret The simplest explanation for both the resorption of the the presence of multiple sanidine phenocryst populations feldspar and quartz cores and the subsequent chemical and and variable zoning patterns of quartz as sampling both the thermal profiles recorded in their rims is a replenishment resident (pre-eruptive) LCT-B magma and the rejuvenation of more juvenile magma into the preexisting mush (e.g., material itself. Ginibre et al. 2004; Stelten et al. 2013; Bachmann et al. 2014; Till et al. 2015). Of the possible mechanisms that can produce the observed core to rim orthoclase variation in Phenocryst record in summary LCT-B sanidine, we favor scenario A/1 where LCT-B bulk system first experiences the addition of rejuvenation material In summary, we interpret the compositions and textures of leading to heating and resorption textures as well as enrich- the LCT-B feldspar and quartz phenocrysts as suggesting a ment in magmaphile elements, followed by decompression composite history for the LCT-B pre-eruptive system: (1) to catalyze the rim crystallization with the observed thermal crystallization of quartz cores and Population 1 and 2 sani- and chemical characteristics. dine cores in a mush stored at 80–150 MPa (which possibly Invoking rejuvenation to explain LCT-B phenocrysts is experienced more than one rejuvenation event indicated by consistent with related Yellowstone studies that find the complexly zoned crystal interiors), (2) rejuvenation of a injection of juvenile magma is an important mechanism more juvenile silicic magma into this mush that provides in the formation of Yellowstone lavas erupted after LCT the impetus for the resorption of the quartz and Population 1 (Stelten et al. 2013, 2015, 2017; Till et al. 2015). In addi- and 2 sanidine cores and enriches the melt in Ba, Sr, Ti, and tion, LCT-B zircons exhibit crystal faces with less-evolved (3) decompression-related growth of quartz and Population 1 compositions and are interpreted to record episodic heat- and 2 sanidine rims which incorporate the higher abundance ing and/or new magma input (Matthews et al. 2015). The magmaphile elements from the rejuvenation material.

1 3 32 Page 14 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

Likely eruption trigger nature of an explosive eruption but also when comparing to other published rates acquired from diffusion chronom- In order to trigger an eruption, sufficient overpressure needs etry. For example, Gualda and Sutton (2016) calculated to be achieved such that dike propagation can occur (e.g., decompression occurred within 1 year or less leading to the Jellinek and DePaolo 2003; Degruyter et al. 2016). The eruption of the Bishop Tuff and Kelley and Wartho (2000) simplest explanation to produce the necessary overpres- obtained timescales of hours to days for kimberlite ascent sure, as suggested by the phenocryst record, is the addition from mantle to surface. Instead, ascent rates calculated for of material during the rejuvenation event. In this scenario, LCT-B are more comparable with ascent rates acquired the resident LCT-B reservoir and surrounding crust are from U-series disequilibria ranging from 10 to 1000 m yr−1 unable to accommodate the additional volume of rejuvena- (Stracke et al. 2006; Rubin et al. 2005). Alternatively, our tion material. The recharge of magma, if it occurs rapidly shorter timescales from crystal growth rate estimates suggest and for a prolonged duration, has been shown in thermome- ascent rates of ~ 480–7700 m yr−1. This faster ascent rate chanical models to be capable of producing the necessary is more comparable to those calculated for other explosive, overpressures to trigger an eruption (e.g., 0.05 km3 year−1 large silicic eruptions and may represent ascent rates aided for 60 years; Degruyter et al. 2016). It is worth noting that, by the additional driving force of second boiling. the ability to trigger an eruption in these models is heavily In summary, we argue that rejuvenation of the magmatic dependent on the degree of crystallinity of the magmatic reservoir is the most likely eruption trigger for the LCT-B system prior to rejuvenation, where lower crystallinity mag- eruption, as it is the simplest process recorded in the feldspar mas are more mobile and therefore more likely to ascend, and quartz phenocryst textures and compositions that could compared to higher crystallinity magmas (e.g., Lejeune and also produce the overpressure to drive the ascent of magma. Richet 1995; Gualda et al. 2012; Degruyter et al. 2016; Ber- Our calculations suggest second boiling during the subse- gantz et al. 2017). If we rely on the extent of crystallization quent ascent and crystallization of the LCT-B magma may predicted by rhyolite-MELTS for the LCT-B whole rock have supplied an additional driving force for the explosive composition (Fig. 7), the system was ~ 15% crystalline at nature of the eruption and promoted faster ascent rates fol- the temperatures and pressures predicted by phenocryst rim lowing rejuvenation. thermobarometry (relevant to the syn- or post-rejuvenation state), far above the predicted crystal content of “rheological Comparison to other caldera‑forming eruptions lock-up” (i.e., 40–60 vol% crystalline; Huber et al. 2012). This suggests LCT-B magma may have been primed for When examining petrologic studies on other caldera-forming remobilization and ascent following the rejuvenation (e.g., eruptions, Flaherty et al. (2018), Sigurdsson et al. (1985) Bachmann and Bergantz 2003). and Morgan et al. (2006) investigated zoned phenocrysts During decompression, it is likely that the rapid crys- to invoke magma injection and subsequent mixing during tallization of LCT-B phenocryst rims leads to significant the few centuries to years prior to the Minoan eruption volatile release (i.e., second boiling), a process which could of Santorini volcano and the ad 79 eruption of Vesuvius. have contributed to the overpressure required for an explo- Alternatively, Gualda and Sutton (2016) and Allan et al. sive eruption and/or increased the magmatic ascent rates (2013) both find no evidence for rejuvenation as a triggering following rejuvenation. The volatile contents constrained by mechanism for the eruption of the Bishop Tuff at Long Val- quartz-hosted melt inclusions (3.4 wt% H2O, Gansecki 1998; ley and the Oruanui eruption at Taupo Volcano, but rather 4 wt% H2O, Befus et al. 2018) combined with the rim crys- invoke decompression as the driver for crystal growth, and tallization pressures, constrained by various geobarometers on relatively short timescales too [1 year (Gualda and Sutton (80–150 MPa; Gualda et al. 2012; Gualda and Ghiorso 2014) 2016) and ≤ 1600 years (Allan et al. 2013)]. Our analyses of predict LCT-B would have experienced a significant positive LCT-B suggest a mixture of these two sets of studies, with volume change (5–15 vol%) associated with vapor exsolu- timescales on the short end of those invoking rejuvenation tion (Sisson and Bacon 1999). Rhyolite-MELTS predicts to eruption timescales despite its significant size, and on the that LCT-B magma was volatile-saturated at the recorded long end of those simply invoking decompression-related rim temperatures, consistent with this possible scenario crystal growth. It should be noted, however, the prior compa- (e.g., Degruyter et al. 2016; Fig. 7). rable diffusion chronometry studies commonly examine only Using the timescales constrained by the feldspar and a single element-mineral pair to calculate a timescale (e.g., quartz growth (years or less) and those constrained by trace Wilson et al. 2006; Morgan et al. 2006; Wark et al. 2007; element diffusion (less than a few decades), we can calculate Allan et al. 2013; Chamberlain et al. 2014; Gualda and Sut- an ascent rate for the LCT-B magma. If we use a conserva- ton 2016; Pamukcu et al. 2016; Rubin et al. 2017; Flaherty tive timescale of 20 years, this equates to an ascent rate of et al. 2018), rather than multiple elements with different dif- ~ 60–450 m yr−1, which is relatively slow, not only for the fusivities, and thus may overlook profiles that are a product

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 15 of 18 32 of crystal growth in an evolving magma composition rather Acknowledgements Thanks to the National Park Service for the scien- than solely diffusion. Therefore, these timescales should be tific permit (YELL-2015-SCI-6078) that made this research possible. We thank Matt Coble (Stanford-U.S. Geological Survey SHRIMP-RG viewed as maxima. When the LCT-B Ba in sanidine concen- lab) for assistance during CL imaging as well as Axel Wittman and tration profiles are treated as resulting from diffusion alone, Maitrayee Bose (Arizona State University) for assistance with the they yield timescales of thousands of years, whereas looking electron microprobe and NanoSIMS analyses respectively. Thank you at elements with multiple diffusivities as done here reveals a to Mark Ghiorso for helpful discussions regarding rhyolite-MELTS modeling. The authors would also like to thank Tim Druitt and Mary more complex story dictated by sanidine growth and time- Reid for their thoughtful and constructive comments that improved scales of decades or less. this manuscript. This study was supported by an NSF CAREER Grant Magmatic recharge or rejuvenation appears to be a com- EAR-1654584 to Till. mon precursor to these large volcanic eruptions and LCT-B appears to represent another such example. Continuing to conduct similar studies on other large eruptions in the Yel- References lowstone system, as well as other caldera-forming eruptions worldwide, is critical for building up statistically robust Allan ASR, Wilson CJN, Millet MA, Wysoczanski RJ (2012) The datasets of the specific triggering mechanism, in addition to invisible hand: tectonic triggering and modulation of a rhyolitic supereruption. Geology 40:563–566. https://doi.org/10.1130/ characteristic trigger-to-eruption timescales for these erup- G32969.1 tions. For example, there is some suggestion that deep long- Allan ASR, Morgan DJ, Wilson CJN, Millet MA (2013) From mush period earthquakes may be associated with fluid exsolution to eruption in centuries: assembly of the super-sized Oruanui of magmatic systems (e.g., Nichols et al. 2011; Sisson et al. magma body. Contrib Mineral Petrol 166(1):143–164 Bachmann O, Bergantz GW (2003) Rejuvenation of the Fish Canyon 2017; Cassidy et al. 2016, 2017), and further work to sub- magma body: a window into the evolution of large-volume silicic stantiate this link, in concert with studies such as this, could magma systems. Geology 31(9):789 facilitate mitigating hazard for future caldera-forming erup- Bachmann O, Huber C (2016) Silicic magma reservoirs in the Earth’s tions. Such datasets can then have the potential to inform our crust. Am Mineral 101:2377–2404. https://doi.org/10.2138/ am-2016-5675 approaches to volcano monitoring, and our hazard assess- Bachmann O, Deering CD, Lipman PW, Plummer C (2014) Building ments of active silicic volcanic systems. zoned ignimbrites by recycling silicic cumulates: insight from the 1,000 km3 Carpenter Ridge Tuff, CO. Contrib Mineral Petrol 167:1–13. https://doi.org/10.1007/s00410-014-1025-3 Befus KS, Bruyere RH, Manga M (2018) Lava Creek Tuff Love. Gold- schmidt Abstracts 162 Summary and conclusions Bergantz GW, Schleicher JM, Burgisser A (2017) On the kinematics and dynamics of crystal-rich systems. J Geophys Res Solid Earth Our petrologic investigation of bulk fallout ash and phe- 122(8):6131–6159 Bindeman IN, Valley JW (2001) Low-d18O rhyolites from Yellow- nocryst rims from Yellowstone’s most recent supererup- stone: magmatic evolution based on analyses of zircon and indi- tion (ca. 631 ka) reveals at least one rejuvenation event in vidual phenocrysts. J Petrol 42:1491–1517 which more juvenile magma was introduced to a reservoir Blundy JD, Wood BJ (1991) Crystal-chemical controls on the partition- of preexisting mush, leading to decompression and phe- ing of Sr and Ba between plagioclase feldspar, silicate melts, and hydrothermal solutions. Geochem Cosmochim Acta 55:193–209 nocryst rim growth within a couple of decades or less prior Blundy J, Cashman K, Humphreys M (2006) Magma heating by to the eruption, similar to other studies of caldera systems decompression-driven crystallization beneath andesite volca- (e.g., Morgan et al. 2006; Druitt et al. 2012; Saunders et al. noes. Nat Lett 443:76–80. https://doi.org/10.1038/nature05100 2012). We interpret multiple zoning patterns and popula- Blundy J, Cashman KV, Rust A, Witham F (2010) A case for CO2-rich arc magmas. Earth Planet Sci Lett 290:289–301. https://doi. tions of LCT-B sanidine and quartz as the result of mixing org/10.1016/j.epsl.2009.12.013 the resident LCT-B magma with rejuvenation material itself. Cabaniss HE, Gregg PM, Grosfils EB (2018) The role of tectonic stress Although it is difficult to unambiguously identify the mecha- in triggering large silicic caldera eruptions. Geophys Res Lett nism that generated the overpressure to drive the eruption, 45:3889–3895. https://doi.org/10.1029/2018GL077393 Calzolaio M, Arzilli F, Carroll MR (2010) Growth rate of this work points toward a rejuvenation event producing the alkali feldspars in decompression-induced crystalliza- initial overpressure, and subsequent second boiling during tion experiments in a trachytic melt of the Phlegraean Fields the ascent and crystallization of the LCT-B magma likely (Napoli, Italy). Eur J Mineral 22:485–493. https://doi. supplying additional overpressure and faster ascent rates org/10.1127/0935-1221/2010/0022-2012 Cashman KV, Sparks RSJ (2013) How volcanoes work: a 25 year per- to produce the extremely explosive nature of the eruption. spective. GSA Bull 125:664–690. https://doi.org/10.1130/b3072 Thus, this study informs the timing and triggering mecha- 0.1 nism of a supereruption at Yellowstone Caldera, and moti- Cassidy M, Castro JM, Helo C, Troll VR, Deegan FM, Muir D, Mueller vates continued study of the physical mechanism capable of SP (2016) Volatile dilution during magma injections and impli- 3 cations for volcano explosivity. Geology 44(12):1027–1030 mobilizing > 1000 km of material in such a short amount Cassidy M, Castro J, Helo C, Ebmeier S, Watt S (2017) Combining of time. experimental petrology with InSAR deformation constraints on

1 3 32 Page 16 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

the magmatic system prior to recent eruptions at Kelud volcano, Gregg PM, Gros EB, De Silva SL (2015) Catastrophic caldera-forming Indonesia. EGU Gen Assembly Conf Abstr 19:9156 eruptions II: the subordinate role of magma buoyancy as an erup- Chamberlain KJ, Morgan DJ, Wilson CJN (2014) Timescales of mixing tion trigger. J Volcanol 305:100–113. https://doi.org/10.1016/j. and mobilisation in the Bishop Tuff magma body: perspectives jvolgeores.2015.09.022 from diffusion chronometry. Contrib Mineral Petrol 168:1–24. Gualda GA, Ghiorso MS (2014) Phase-equilibrium geobarometers for https://doi.org/10.1007/s00410-014-1034-2 silicic rocks based on rhyolite-MELTS. Part 1: Principles, pro- Cherniak DJ (1996) Strontium diffusion in sanidine and albite, and cedures, and evaluation of the method. Contrib Mineral Petrol general comments on strontium diffusion in alkali feldspars. Geo- 168(1):1033 chim Cosmochim Acta 60:5037–5043. https://doi.org/10.1016/ Gualda GAR, Sutton SR (2016) The year leading to a supereruption. S0016-7037(96)00293-1 PLoS One 11:1–18. https://doi.org/10.1371/journal.pone.01592 Cherniak DJ (2002) Ba diffusion in feldspar. Geochim Cosmochim Acta 00 66:1641–1650. https://doi.org/10.1016/S0016 -7037(01)00866 -3 Gualda GAR, Ghiorso MS, Lemons RV, Carley TL (2012) Rhyolite- Cherniak DJ, Watson EB, Wark DA (2007) Ti diffusion in quartz. Chem MELTS: a modified calibration of MELTS optimized for silica- Geol 236:65–74. https://doi.org/10.1016/j.chemg eo.2006.09.001 rich, fluid-bearing magmatic systems. J Petrol 53:875–890. https Christiansen RL (2001) The quaternary and pliocene Yellowstone Pla- ://doi.org/10.1093/petrology/egr080 teau Volcanic Field of Wyoming, Idaho, and Montana. USGS Guo J, Green TH (1989) Barium partitioning between alkali feldspar Professional Papers. No. 729-G and silicate liquid at high temperature and pressure. Contrib Min- Costa F, Morgan D, Dosseto A, Turner S, Van Orman J (2011) Time eral Petrol 102(3):328–335 constraints from chemical equilibration in magmatic crystals. Hawkesworth C, George R, Turner S, Zellmer G (2004) Time scales Timescales of magmatic processes: from core to atmosphere. of magmatic processes. Earth Planet Sci Lett 218:1–16. https:// Wiley, Chichester, pp 125–159 doi.org/10.1016/S0012-821X(03)00634-4 Couch S, Harford CL, Sparks RSJ, Carroll MR (2003) Experimental Hayden LA, Watson EB (2007) Rutile saturation in hydrous siliceous constraints on the conditions of formation of highly calcic plagio- melts and its bearing on Ti-thermometry of quartz and zircon. clase microlites at the Soufriere Hills Volcano, Montserrat. J Pet- Earth Planet Sci Lett 258:561–568. https://doi.org/10.1016/j. rol 44:1455–1475. https://doi.org/10.1093/petrology/44.8.1455 epsl.2007.04.020 Degruyter W, Huber C, Bachmann O et al (2016) Magma reservoir Hildreth W (1981) Gradients in silicic magma chambers: implica- response to transient recharge events: the case of Santorini tions for lithospheric magmatism. J Geophys Res Solid Earth volcano (Greece). Geology 44:23–26. https://doi.org/10.1130/ 86(B11):10153–10192 G37333.1 Hildreth W, Christiansen RL, O’Neil JR (1984) Catastrophic isotopic Druitt TH, Costa F, Deloule E, Dungan M, Scaillet B (2012) Decadal modification of rhyolitic magma at times of caldera subsidence, to monthly timescales of magma transfer and reservoir growth Yellowstone Plateau volcanic field. J Geophys Res Solid Earth at a caldera volcano. Nature 482:77–80. https://doi.org/10.1038/ 89(B10):8339–8369 nature10706 Hildreth W, Halliday AN, Robert L (1991) Isotopic and chemical evi- Eaton GP, Christiansen RL, Iyer HM, Pitt AM, Mabey DR, Blank HR, dence concerning the genesis and contamination of basaltic and Zietz I, Gettings ME (1975) Magma Beneath Yellowstone. Sci- rhyolitic magma beneath the Yellowstone Plateau Volcanic Field. ence 188:787–796 J Petrol 32:63–138 Farrell J, Smith R, Husen S, Diehl T (2014) Tomography from 26 years Huang R, Audétat A (2012) The titanium-in-quartz (TitaniQ) ther- of seismicicty revealing that the spatial extent of the Yellowstone mobarometer: a critical examination and re-calibration. Geo- crustal reservoir extends well beyond the Yellowstone caldera. chim Cosmochim Acta 84:75–89. https://doi.org/10.1016/j. Geophys Res Lett 41:3068–3073. https://doi.org/10.1002/2014G gca.2012.01.009 L059588 Huber C, Bachmann O, Dufek J (2012) Crystal-poor versus crystal- Fenn PM (1977) The nucleation and growth of alkali feldspars from rich ignimbrites: a competition between stirring and reactivation. hydrous melts. Can Mineral 15:135–161 Geology 40:115–118. https://doi.org/10.1130/G32425.1 Flaherty T, Druitt TH, Tuffen MD et al (2018) Multiple timescale Humphreys MCS, Edmonds M, Klöcking MS (2016) The validity of constraints for high-flux magma chamber assembly prior to the plagioclase-melt geothermometry for degassing-driven magma Late Bronze Age eruption of Santorini (Greece). Contrib Mineral crystallisation. Am Mineral 101:769–779 Petrol 173:75. https://doi.org/10.1007/s00410-018-1490-1 Iacovino K (2015) Linking subsurface to surface degassing at active Gansecki CA (1998) 40Ar/39Ar geochronology and pre-eruptive geo- volcanoes: a thermodynamic model with applications to Ere- chemistry of the Yellowstone Plateau volcanic field rhyolites. bus volcano. Earth Planet Sci Lett 431(59):74. https://doi. Unpublished doctoral dissertation, Stanford University, p 212 org/10.1016/j.epsl.2015.09.016 Gansecki CA, Mahood GA, McWilliams M (1998) New ages for the Icenhower J, London D (1996) Experimental partitioning of Rb, Cs, climactic eruptions at Yellowstone: single-crystal 40Ar/39Ar dat- Sr, and Ba between alkali feldspar and peraluminous melt. Am ing identifies contamination. Geology 26(343):346. https://doi. Mineral 81:719–734. https://doi.org/10.2138/am-1996-5-619 org/10.1130/0091-7613 Jellinek AM, DePaolo DJ (2003) A model for the origin of large Gardner JE, Befus KS, Gualda GAR, Ghiorso MS (2014) Experimental silicic magma chambers: precursors of caldera-forming erup- constraints on rhyolite-MELTS and the Late Bishop Tuff magma tions. Bull Volcanol 65:363–381. https://doi.org/10.1007/s0044 body. Contrib Mineral Petrol 168:1–14. https://doi.org/10.1007/ 5-003-0277-y s00410-014-1051-1 Jicha BR, Singer BS, Sobol P (2016) Re-evaluation of the ages of 40 Ghiorso MS, Gualda GAR (2015) An H 2O–CO2 mixed fluid saturation Ar/39 Ar sanidine standards and supereruptions in the western model compatible with rhyolite-MELTS. Contrib Mineral Petrol US using a Noblesse multi-collector mass spectrometer. Chem 169:1–30. https://doi.org/10.1007/s00410-015-1141-8 Geol 431:54–66 Ginibre C, Wörner G, Kronz A (2004) Structure and dynamics of the Kelley SP, Wartho JA (2000) Rapid kimberlite ascent and the sig- Laacher See magma chamber (Eifel, Germany) from major and nificance of Ar–Ar ages in xenolith phlogopites. Science trace element zoning in sanidine: a cathodoluminescence and 289(5479):609–611 electron microprobe study. J Petrol 45:2197–2223. https://doi. Leeman W, Vicenzi E, Macrae C et al (2008) Systematics of cathodo- org/10.1093/petrology/egh053 luminescence and trace element compositional zoning in

1 3 Contributions to Mineralogy and Petrology (2019) 174:32 Page 17 of 18 32

natural quartz from volcanic rocks: Ti mapping in quartz. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to Microsc Microanal 18:1322–1341. https://doi.org/10.1017/ ImageJ: 25 years of image analysis. Nat Methods 9(7):671–675 S1431927608088995 Sigurdsson H, Carey S, Cornell W, Pescatore T (1985) The eruption Leeman WP, MacRae CM, Wilson NC, Torpy A, Lee CT, Student of Vesuvius in A.D. 79. Natl Geogr Res 1(3):332–387 JJ, Thomas JB, Vicenzi EP (2012) A study of cathodolumi- Sisson TW, Bacon CR (1999) Gas-driven filter pressing in mag- nescence and trace element compositional zoning in natu- mas. Geology 27:613–616. https://doi.org/10.1130/0091- ral quartz from volcanic rocks: mapping titanium content in 7613(1999)027%3c0613:GDFPIM%3e2.3.CO quartz. Microsc Microanal 18(6):1322–1341 Sisson T, Power J, Haney M (2017) The deep-crustal roots of arc Lejeune AM, Richet P (1995) Rheology of crystal-bearing silicate stratovolcanoes illuminated by long-period seismicity caused melts: an experimental study at high viscosities. J Geophys Res by degassing and fluid-enhanced wallrock embrittlement. IAV- Solid Earth 100(B3):4215–4229 CEI General Assembly Submission 920 Long PE (1978) Experimental determination of partition coefficients Smith RB, Braile LW (1994) The Yellowstone hotspot. J Volcanol for Rb, Sr, and Ba between alkali feldspar and silicate liquid. Geotherm Res 61:121–187. https://doi.org/10.1016/0377- Geochim Cosmochim Acta 42:833–846 0273(94)90002-7 Mark DF, Renne PR, Dymock RC, Smith VC, Simon JI, Morgan LE, Smith J, Brown WL (1988) Feldspar minerals: crystal structures, Pearce NJ (2017) High-precision 40Ar/39Ar dating of Pleisto- physical, chemical and microtextural properties. Springer, cene tuffs and temporal anchoring of the Matuyama-Brunhes Berlin, p 828 boundary. Quat Geochron 39:1–23 Smith RB, Jordan M, Steinberger B et al (2009) Geodynamics of the Matthews NE, Huber C, Pyle DM, Smith VC (2012) Timescales Yellowstone hotspot and mantle plume: seismic and GPS imag- of magma recharge and reactivation of large silicic systems ing, kinematics, and mantle flow. J Volcanol Geotherm Res from Ti diffusion in quartz. J Petrol 53:1385–1416. https://doi. 188:26–56. https://doi.org/10.1016/j.jvolgeores.2009.08.020 org/10.1093/petrology/egs020 Sparks S (2005) Super-eruptions: global effects and future threats, Matthews N, Vazquez J, Calvert A (2015) Age of the Lava Creek report of a Geological Society of London Working Group. supereruption and magma chamber assembly at Yellowstone Geological Society, London based on 40Ar/39Ar and U–Pb dating of sanidine and zircon Stelten ME, Cooper KM, Vazquez JA et al (2013) Magma mixing and crystals. Geochem Geophys Geosyst 16:1–21. https://doi. the generation of isotopically juvenile silicic magma at Yel- org/10.1002/2015GC005881 lowstone caldera inferred from coupling 238 U–230 Th ages Morgan DJ, Blake S, Rogers NM et al (2006) Magma chamber with trace elements and Hf and O isotopes in zircon and Pb recharge at Vesuvius in the century prior to the eruption of isotopes in sanidine. Contrib Mineral Petrol 166(2):587–613. A.D. 79. Geology 34:845–848. https://doi.org/10.1130/G2260 https://doi.org/10.1007/s00410-013-0893-2 4.1 Stelten ME, Cooper KM, Vazquez JA et al (2015) Mechanisms and Müller A, Wiedenbeck M, van den Kerkhof AM et al (2003) Trace timescales of generating eruptible rhyolitic magmas at Yel- elements in quartz—a combined electron microprobe, second- lowstone Caldera from Zircon and sanidine geochronology and ary ion mass spectrometry, laser-ablation ICP-MS, and cathodo- geochemistry. J Petrol 56:1607–1642. https://doi.org/10.1093/ luminescence study. Eur J Mineral 15:747–763. https://doi. petrology/egv047 org/10.1127/09351221/2003/0015-0747 Stelten ME, Cooper KM, Wimpenny JB et al (2017) The role of man- Nichols ML, Malone SD, Moran SC, Thelen WA, Vidale JE (2011) tle-derived magmas in the isotopic evolution of Yellowstone’s Deep long-period earthquakes beneath Washington and Oregon magmatic system. Geochem Geophys Geosyst 18:1350–1365. volcanoes. J Volcanol Geotherm Res 200(3–4):116–128 https://doi.org/10.1002/2016GC006664 Oppenheimer C, Moretti R, Kyle PR et al (2011) Mantle to surface Stracke A, Bourdon B, McKenzie D (2006) Melt extraction in the degassing of alkalic magmas at Erebus volcano, Antarctica. Earth’s mantle: constraints from U-Th–Pa–Ra studies in oce- Earth Planet Sci Lett 306:261–271. https://doi.org/10.1016/j. anic basalts. Earth Planet Sci Lett 244:97–112 epsl.2011.04.005 Streck MJ (2008) Mineral textures and zoning as evidence for open Pamukcu AS, Ghiorso MS, Gualda GAR (2016) High-Ti, bright-CL system processes. Rev Mineral Geochem 69:595–622. https:// rims in volcanic quartz: a result of very rapid growth. Con- doi.org/10.2138/rmg.2008.69.15 trib Mineral Petrol 171:1–9. https://doi.org/10.1007/s0041 Swanson SE (1974) Phase equilibria and crystal growth in granodior- 0-016-1317-x itic and related systems with H2O + CO2. Ph.D. thesis, Stanford Putirka KD (2008) Thermometers and barometers for volcanic sys- University tems. Rev Mineral Geochem 69:61–120. https://doi.org/10.2138/ Swanson SE (1977) Relation of nucleation and crystal-growth rate to rmg.2008.69.3 the development of granitic textures. Am Mineral 62:966–978 Ren M (2004) Partitioning of Sr, Ba, Rb, Y and LREE between Thomas JB, Watson EB, Spear FS et al (2010) TitaniQ under pres- alkali feldspar and peraluminous silicic magma. Am Mineral sure: the effect of pressure and temperature on the solubility 89(8–9):1290–1303 of Ti in quartz. Contrib Mineral Petrol 160:743–759. https:// Rubin KH, van der Zander I, Smith MC, Bergmanis EC (2005) Mini- doi.org/10.1007/s00410-010-0505-3 mum speed limit for ocean ridge magmatism from 210Pb–226Ra– Till CB, Vazquez JA, Boyce JW (2015) Months between rejuvena- 230Th disequilibria. Nature 437:534–538 tion and volcanic eruption at Yellowstone caldera, Wyoming. Rubin AE, Cooper KM, Till CB et al (2017) Rapid cooling and cold Geology. https://doi.org/10.1130/g36862 storage in a silicic magma reservoir recorded in individual crys- Turner S, Costa F (2007) Measuring timescales of magmatic evolu- tals. Science 356:1154–1156. https://doi.org/10.1126/scien tion. Elements 3:267–272 ce.aam8720 Vazquez JA, Kyriazis SF, Reid MR et al (2009) Thermochemical Saunders K, Blundy J, Dohmen R, Cashman K (2012) Link- evolution of young rhyolites at Yellowstone: evidence for a ing petrology and seismology at an active volcano. Science cooling but periodically replenished postcaldera magma res- 336(6084):1023–1027 ervoir. J Volcanol Geotherm Res 188:186–196. https://doi. Schleicher JM, Bergantz GW (2017) The mechanics and temporal evo- org/10.1016/j.jvolgeores.2008.11.030 lution of an open-system magmatic intrusion into a crystal-rich magma. J Petrol 58(6):1059–1072

1 3 32 Page 18 of 18 Contributions to Mineralogy and Petrology (2019) 174:32

Wallace G, Bergantz GW (2002) Wavelet-based correlation (WBC) of characteristics and evacuation of a large rhyolitic magma body. J zoned crystal populations and magma mixing. Earth Planet Sci Petrol 47:35–69. https://doi.org/10.1093/petrology/egi066 Lett 202:133–145 Wilson CJN, Stelten ME, Lowenstern JB (2018) Contrasting perspec- Wallace PJ, Anderson AT et al (1995) Quantification of pre-eruptive tives on the Lava Creek Tuff eruption, Yellowstone, from new exsolved gas contents in silicic magmas. Nature 377:612–616 U-Pb and 40Ar/39Ar age determinations. Bull Volcanol 80(6):53 Wark DA, Watson EB (2006) TitaniQ: a titanium-in-quartz geother- Wotzlaw JF, Bindeman IN, Stern RA et al (2015) Rapid heterogene- mometer. Contrib Mineral Petrol 152:743–754. https://doi. ous assembly of multiple magma reservoirs prior to Yellowstone org/10.1007/s00410-006-0132-1 supereruptions. Sci Rep 5:1–10. https://doi.org/10.1038/srep1 Wark DA, Hildreth W, Spear FS et al (2007) Pre-eruption recharge 4026 of the Bishop magma system. Geology 35:235–238. https://doi. Zellmer GF, Clavero JE (2006) Using trace element correlation patterns org/10.1130/G23316A.1 to decipher a sanidine crystal growth chronology: an example Watts KE, Bindeman IN, Schmitt AK (2012) Crystal scale anatomy from Taapaca volcano, Central Andes. J Volcanol Geotherm Res of a dying supervolcano: an isotope and geochronology study 156:291–301. https://doi.org/10.1016/j.jvolgeores.2006.03.004 of individual phenocrysts from voluminous rhyolites of the Yel- Zhang Y (2008) Geochemical kinetics. Princeton University Press, lowstone caldera. Contrib Mineral Petrol 164(1):45–67 Princeton, p 656 Wilke S, Holtz F, Neave DA, Almeev R (2017) The effect of anorthite content and water on quartz-feldspar cotectic compositions in Publisher’s Note Springer Nature remains neutral with regard to the rhyolitic system and implications for geobarometry. J Petrol jurisdictional claims in published maps and institutional affiliations. 58:789–818. https://doi.org/10.1093/petrology/egx034 Wilson CJN, Blake S, Charlier BLA, Sutton AN (2006) The 26.5 ka Oruanui Eruption, Taupo Volcano, New Zealand: development,

1 3